[Discuss] Cold Boot Attacks on Encryption Keys

[tmetro+blu at gmail.com (Tom Metro)]

Bill Bogstad wrote:
> Cold Boot Attacks on Encryption Keys

If the machine is currently running, I suspect you can accomplish much
the same end result without the complication of cryogenics by simply
attaching a bus analyzer to the memory bus (physically doing that may be
challenging with modern CPU packages and motherboards). An appropriately
designed PCI-X card that uses DMA to dump the memory would be another
approach. Either of these could be done with the machine powered.


> So you need to both cut the power and prevent physical access for a
> few minutes.

I've mentioned this before, but a mercury switch, or other forms of trip
switches, are one way you address this.

The scenario is that you have strongly encrypted data on disk,
decryption keys in memory, an OS configured so that it doesn't do
something stupid, like write the keys to unencrypted swap space, and an
OS hardened enough that physical access to the machine seems like the
easier attack vector.

But then the scenario starts to get a bit more far fetched. The people
seizing your server apparently already know or suspect you are using
full disk encryption, and your data is valuable enough to warrant
bringing in people skilled enough to hot jumper your machine to a
portable power source before moving it back to a lab where the RAM can
be frozen and dumped.

In any case, as soon as the machine is moved or a cover opened, a trip
switch cuts power internally. If they weren't expecting this, you've
increased your chances that all or most of your key will be corrupted by
the time they get some freon on your RAM.

 -Tom

-- 
Tom Metro
Venture Logic, Newton, MA, USA
"Enterprise solutions through open source."
Professional Profile: http://tmetro.venturelogic.com/